Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!apple!bbn!bbn.com!denbeste From: denbeste@bbn.com (Steven Den Beste) Newsgroups: comp.sys.amiga Subject: Re: Benchmarks (was: Re: A2620) Summary: I used to know some of the guys that designed the 8086 and THEY hated it, too. Message-ID: <41426@bbn.COM> Date: 14 Jun 89 17:02:14 GMT References: <780@corpane.UUCP> <7084@cbmvax.UUCP> <25428@agate.BERKELEY.EDU> <6607@dayton.UUCP> Sender: news@bbn.COM Distribution: na Lines: 59 In article <6607@dayton.UUCP>, joe@dayton.UUCP (Joseph P. Larson) writes: > > Intel might make fast or reliable chips. But they weren't designed by > anyone with experience with real assembly languages. (Designers were probably > ex Z-80 or 6502 programmers.) > The 8086 was designed at Intel simultaneously with the 432. The 8086 was expected to be a transition product for their 8-bit users (remember the 8085?) and wasn't intended to last long. Intel had placed its hopes on the 432. The designers were given the following command: Make it assembly source-code compatible with the 8085. That's the reason for the gothic non-orthogonal register architecture we all hate so much. The designers hated it too, but they couldn't do anything about it. After designing the architecture, they came up with a macro package which allowed one to run 8085 ASM on an 8086. But they gritted their teeth and did it anyway. After all, it wasn't going to be around all that long. Anyone out there remember the 432? Definitely a processor ahead of its time. It was designed to be an Ada engine, but you had to hook half a meg of RAM onto it just for its registers, even before you talked about code or data. This was back in the days when a 64K DRAM cost $20 each. In any case, the 432 bombed badly. ...and then IBM came along and bought 30% of Intel and decided to use the 8088 in the PC - and suddenly the 8086 was the hope for Intel's future. Their designers have been living with the 8086's illegitimate parentage ever since as they move up the ladder of performance. Motorola, of course, did it the right way: They designed a 32-bit architecture and then implemented it as a 16-bit data path machine. That meant they could build a 32-bit data path machine later with minimal external change. Generally speaking, I've found that technical people divide into the Intel-philes and the Motorola-philes. The correlation isn't 1.0, but I've found that more often than not an Intel-phile is an EE and a Motorola-phile is a programmer. The Intel chips have always been more friendly to EE's, and the 8086 is no exception. There the designers didn't have to emulate anything, and they did a nice job. Ironically, here Motorola got trapped. Their designers had to make the 68000 work with the 8-bit peripheral chips from the 6800/6809, because when the 68000 was released its own family of chips wasn't available yet. As a result, 68000 hardware tends to be somewhat more baroque, even though it isn't a multiplexed bus. Don't always assume that if you see something that was done badly that it was because the designers were incompetent, stupid or crooked. Often they were laboring under constraints about which you do not know, and you would have done exactly the same thing under those constraints. Steven C. Den Beste, BBN Communications Corp., Cambridge MA denbeste@bbn.com(ARPA/CSNET/UUCP) harvard!bbn.com!denbeste(UUCP)